Available online at wivw.sciencedirect.com Procedia MANUFACTURING ScienceDirect Procedia Manufacturing 54 (2021) 184-190 ELSEVIER ‘wo elsevercomocatprocdia 10th CIRP Sponsored Conference on Digital Enterprise Technologies (DET 2020) — Digital Technologies as Enablers of Industrial Competitiveness and Sustainability Enhance of OEE by hybrid analysis at the automotive semi-automatic assembly lines Péter Dobra, Janos Jésvai? "Doctoral School of Maliditcilinary Engineering Scenes Sobcheny vin Univertiy, Eyton tr 1, 026 Gyr, Hungary Department of Vehicle Manafacturing. Seéhen vin Univer, Egetem i, 9026 Oy, Hungary * Corresponding autor Tel: -36-96503-40 fx: $3696-503-00, E-mail address dobra peters hs Abstract Nowadays, industrial companies are constantly improving their processes and increasing their efficiency in order to achieve higher profits. One fof the most common measures of efficiency atthe semi-automatic assembly lines isthe Overall Equipment Effectiveness (OEE) as a Key Performance Indicator (KPI). The goal of measurement of OFE is to improve the effectiveness of machines or production lines with minimal invesiments. The article presents 8 way t inerease this indicator using sa own hybrid analysis, Firstly, a Titeratre review demonstates, scientific relevance. Secondly, a hybrid analysis is introduced for improving te efficiency. The effectiveness of the analysis is demonsrated by a practical exa fusing data mining and line balancing, during wi ‘machine combined with human analysis provides this outstanding result. Hybrid here was a 60% improvement in yield, Pater recognitions by alysis ean be used not only on assembly lines, but also on individual machines where a fot of data is generated and the eycle time or takt time neods to be reduced for higher yields © 2021 The Authors. Published by Elsevier B.V. ‘This is an open access article under the CC BY-NC-ND license (}ip/ereativecommons.org/licenses!by-ne-nd!4.0) Peer-review under responsibility of the scientific committee ofthe 1th CIRP Sponsored Conference on Digital Enterprise Technologies (DET 2020) — Digital Technologies as Enablers of Industrial Competitiveness and Sustainability Keywords: OBE; data mining: line balancing: assembly ine 1. Introduction The participants of industry, thus manufacturing companies are under increasing pressure to improve productivity and profitability day by day. Production logistics are among the ‘most important and relevant optimization targets for the task ‘of increase the performance of assembly systems [1]. The ‘enterprises use different Key Performance Indicators (KPI's) in daily operations to follow up their performance and systematically improve the main production and assembly processes. The scientific literature proposes different calculation models for estimating the productivity of a ‘machine or an assembly line [2]. In order to continuously monitor and track the effectiveness the Overall Equipment fectiveness (OEE) indicator is one of the best measurement 2351-9789 © 2021 The Authors. Published by ‘This ie an open access atticle under the CC BY eevier BLY techniques. Nowadays, atthe assembly Tines there are several methods and techniques - eg. lean manufacturing, line balancing and six sigma ~ to enhance the value of OEE, These activities are supported by the environment of Industry 4.0, Smart manufacturing, Intemet of Things (IoT), Big Data, Artificial Imelligence (AI) using softwares and algorithms to analyse real time data, ‘The increasing digitalization of manufacturing industry provides means to automatically acquire and analyse the data although companies do not use data mining methods widely, parily due to the need of database expert, data mining expert ‘and domain expert on the human resources side [3.4.5] Generally, the industrial firms focus on using data for analysis, monitoring and control e.g. production efficiency analysis, process monitoring and quality control [6]. Without INC-ND ieense (http://reativecommons.ong/licenses/by-ne-nd/4.0/) Pect-review under responsibility of the scientific committee of the 10 IRP Sponsored Conference on Digital Enterprise Technologies {DET 2020) — Digital Technologies as Enablers of Industrial Competitiveness and Sustainability 10.1016/3 promt, 2021.07.028Pérer Dobra etal. /Procedia Manufacturing $4 (2021) 184-190 18s using data mining methods, the hidden patterns can not be revealed in manuafacturing-related data [7] This article focuses on the common section of assembly line, data mining and line balancing (Fig. 1.). To enhance OEE based on their connection a hybrid analysis will be presented 2, State of the Art In the modem world, industrial companies are surrounded by so much information and data that they are oflen unable to ‘manage and process them efficiently enough, Manufacturing and assembly operations become more automated and computerized, IT systems and networks are connected to ‘machines, products and operators. The combination of lean production, smart manufacturing and Industry 4.0 offers new ‘opportunities from the side of proactive, reactive and predictive approuch [8]. However, companies are often not aware of opportunities provided by the combination of lean production and digitization [9]. This chapter provides a brief literature review of OFB, line balancing and data mining, 2.1. Overall Equipment Effectiveness (OBE) In the industrial practice there are several Indicators (Is Performance Indicators (PI's) and Key Performance Indicators (KPI's) which shows the effectiveness of individual machines for assembly ines. Overall Equipment Effectiveness is a ‘widely used and most famous performance indicator [10]. It vwas presented by Nakajima when the Total Productive ‘Maintenance (TPM) concept was introduced in 1988 [11]. The basic formula for calculating OBE is written as: OBE =apq (%) w where a - availability [2]; p - performance [%I; 4 - quality (26) EE indicator are characterizedin the following ways: + reliability ofthe production network (12) + valuable tool to unleash the hidden capacity [13] + used to measure equipment productivity [14] # a productivity ratio between real manufacturing and what could be ideally manufectured [14] + benefits at the side of equipment, personnel, process and quality [15] ‘+ identify the real oot of the problems [16] OE is described as @ multi-layered indicetor that can be used in many industries regardless of whether the production is continuous [17] or batch-sized [15]. Ales etal, presented a etsiled calculation in the case of serial, parallel and combined system [18]. Measuring the OEE of an assembly system is essential for different purposes including, scheduling, capacity estimation, human resource allocation, and budgeting, From the original OBE structure during the last three decades many performance measurement techniques and applications were developed such as Overall Asset Effectiveness (OAB), Overall Equipment Effectiveness of @ Manufacturing Line (OEEML), Global Production Effectiveness (GPE), Overall Throughput Effectiveness (OTE), Equipment Performance and Reliability (EPR), Overall Weighting Equipment Effectiveness (WOEE) and Overall Equipment and Quality Cost Loss (OEQCL) [19] Whichever metric is used, it is important that the measures taken are based on trends, 2.2, Significance of line balancing and bottleneck In the individual machines and the assembly lines domain it is well-known, that the productivity of manufacturing is, constraint by throughput bottlenecks. Bottleneck machines or stations are critical and determine the largest sensitivity to the production system performance [20,21], For these reasons, the bottleneck station needs to be analyzed and improved with ifferent detection methods e.g. turning point method, active period method, arrow method, etc. [22] Ina turbulent market environment, the inereasing number of products variety and constant change of the assembly process causes a vatying processing time and demands the rebalancing of assembly lines [23,24]. Trom the le ‘manufacturing side, where the most important goal is to climinate waste, a wide-spreaded and applied method is the line balancing, Line balancing is a technique to achieve balanced state between operators and workloads in order to fulfill the required run rate. Regardless of the life cycle of the product produced on the production line, line balancing is required from time to time in assembly lines [25] ‘Assembly line balancing was introduced by Bryton in 1954. The goal of simple assembly line balancing problem (SALBP) are: ‘© minimalize the number determined cycle time ‘+ minimalize the eycle time besides number of workstations (26) ‘+ assigning operations to workstations along an assembly Tine [27] of workstations besides the ElMaraghy ef al, raised numerous assembly Tine balancing, methods as dynamic programming, integer programming tabu search and combination of a heuristic model and an exact algorithm (28). Manaye compared different line balancing techniques and methods with each other eg. ranked positional ‘weighted method, simulations method, largest candidate rule method and probabilistic techniques [27]. Mura et al186 Pé1er Dobra etal. /Procedia Manufacturing $4 2021) 184-190) presented an algorithm ealled genetic iteration for assembly lines (GenlAl) [29]. Mourtzis et al. introduced and applied in 4 practice a new load balancing and planning method to provide decision support for line balancing [30,31]. Sly presented the practical visual application of Yamazumi chart ‘on an assembly line where value-added and non-added value activites were performed by operators in different ‘workstations (32) 23 Data Mining at the assembly area The growth of the digital universe is mainly due to spreading of sensors and development of new technologies. Data generated automatically by machines results in exponentially increasing amount of digital data [33]. In ‘modern production environment, parts of the products are labelled with barcodes and the workstations of assembly lines, are equipped with numerous sensors and vision systems. Al ‘of devices, equipment, stations and tools are integrated in one network hence more and more data are recorded and each assembly steps can be traced [34,35]. Industrial enterprises, are increasingly using wireless technologies and ‘Manufacturing Execution System (MES) to capture data at all stages of @ product's life cycle [36]. MES can collect shop floor data in digital format and provide an entire real-time database for production control [20,37] ‘Data mining has been used in many fields such as banking, insurance, medicine, marketing, sales and industry [38], Data ‘mining began to use in the field of production in the 1990's, One of the earliest and most common definition of data mining by Frawley is the following: “The non-trivial extraction of implicit, previously unknown, and potentially useful information from data”. [39] At the area of shop floor, data mining is applied in predictive maintenance, fault detection, production, quality assurance, scheduling, process planning and decision support system [40]. Choudhary etal pointed out thatthe one of the advantages of data raining over other techniques is that the required data for analysis can be collected during the normal manufacturing and assembly processes [41] Data mining is a step in the Knowledge Discovery in Databases (KDD) process, and the aim is the extracting mowledge and discovering interesting pattems from large amount of date [42,4344]. Numerous data mining methods, ‘models and techniques applied within manufacturing and assembly environments such as association discovery, classification, regression, clustering, summarization, dependency modelling, change and deviation detection, example-based methods, relational learning models, sequential pattem discovery, temporal modelling and characteristic discovery [5,42,44,45,46,47]. Nevertheless, there is no universal data mining method, One of the best known data mining process is the Industry Standard Process for Data Mining (CRISP-DM) which is independent of both the industrial technology and sector [48], This model was extended with new elements, (echnical understanding, technical realization and technical implementation) and published by Huber et al. as Data Mining Methodology for Engineering (DMME) [49], SEMMA, Kensington Enterprise Data Mining, SolEuNet and DMG among others, established methodologies, developed languages and softwares for standardization [40]. Data mining js supported by many software and extractions tools as Weka [50], SPSS, SAS, Rapidminer, Tanagra, ODM, R, Python, ete 3. Analysis for improve OEE Assembly lines 2s production systems are typical in the ‘mass production of standard produets, Assembly operations of manufactured goods account for over 30% of total production time and 20% of total production cost therefore it deserves special focus in manufacturing companies [28]. An assembly line consists of several workstations that generally have different eycle times, There is a bottleneck on the lines where either the assembly, manufacturing, testing or inspection operations of a product requires the longest time. To ‘maximize the output from an assembly lin, the throughput of the bottleneck needs to be improved. In ease the eyele time in the bottleneck station is reduced and leaving the other factors unchanged (serap rate, ete.) then, it leads to an inerease in the OBE value. 3.1. Hybrid analysis for increasing OBE Cycle time is one of the most important data for the line balancing at an assembly line. At each workstation for each produet the modern systems record this value, which results in fan ever-increasing significant amount of dala at mass production. Manual processing of such amount of data ean not be efficient, but machine processing is not fully capable 10 analyse the data regarding to practical experience and correlations. For example, congestion on the assembly line for technical, quality, logistics or human reasons. The combination of manual analysis and machine pattern recognition is illustrated by the new hybrid analysis (Fig. 2), which aims to enhance OBE value Fig 2 Hye analysis 3.2, Elements ofthe hybrid analysis: + Current assembly line operation The outlined analysis can be used for assembly lines which operate semi-automatically (operator + machine) and the cyele time of each station is not constant, thus the bottleneck ‘may constantly change between certain groups of stations.Pérer Dobra etal. /Procedia Manufacturing $4 2021) 184-190) 187 ‘The reason for the change could be the different levels of ‘operator practice and the different testing, re-measurement and self-calibration time of the stations, product variants, et. ‘+ Measured eyele time at the workstations Cycle time measurement and analysis is performed at bottleneck station using manual time measurement at the ‘motion elements detection by the method of Methods Time ‘Measurement (MTM). For example, measure the time of serew tightening or grabbing 2 part. The operators are the ‘most important and critieal factor which influences the line productivity (13) ‘+ Human analysis 'At engineering level, human analysis focuses on the relationships and interactions between stations. It is advisable to observe and analyze the effeets when one station affeets the cycle time of the other station (e.g,, improper assembly increases the test time, rework operations due to serial errors, tc.). Human analysis is based on human observations and take into consideration the human experiences. ‘+ Data mining based on MES During data mining, the large amount of historical data set is, processed, which determines the unique cycle time of each product for each station. The base of data mining is the MES and from the SQL system the databases are processed, ‘» Patterns recognitions by machine During the analysis by machine, the location of the bottleneck is explored and the smallest eyele time that resulted in the proper (non-scrap) product is assigned. Patterns recognition are performed based on the relationship between cycle time and workstations ‘© Design new balanced line Following manual and machine analysis, line balancing is, performed based on the data and relationships obtained. Line ‘balancing is based on the actual eye times performed at each sation and the motion elements detected by the human with time data, The method allows for a comprehensive analysis, ‘bocause it takes into account all the data generated in the past, and assigns the relationships discovered by man 10 the patterns found by the machine, In practice, the analysis can be applied over several cycles and the validation is by the ‘machine 3.3. Case study at an automotive seat structure semi- ‘automatic assembly line ‘There are many examples in industrial practice of searching for patterns using data mining or performing fine balancing on assembly lines, These techniques used separately probably do not give such 2 good result like applying them combined in several cycles. The efficiency of the hybrid analysis has been demonstrated through an automotive semi-automatic assembly ine ‘The seat structure assembly line contains a total of 15 workstations (Fig. 3). The line has a mixed structure of ‘manual assembly stations, machine assembly stations, measuring and testing stations, The products are moved between stations on pallets. The product is in growth stage of its life cycle, work instructions and Standard Operating Procedure (SOP) are available, the line is partially balanced Fig. 3. Semi-automatic assembly ie ‘The cycle time data of each station is provided by the MES, as well as all other data necessary for the calculation of the OEE (downtime, scrap rate, et.). The MES automatically calculates the OEE value for the given period (hour, shift, day, cte), For data mining one year of production data was used and 2.7 million data rows were collected. ata processing and data mining were performed using the Excel - Analysis ToolPak, As a fist step, the average daily cycle time of each station was determined (Fig. 4.) and cycle time ranges and groups were created using "clustering procedure as supervised machine learning. Fig, 5. shows the related density function, The cycle time ranges are the sections into which the measured cycle times have been assigned. In this case, the best solution was shown by the 25, sec ranges in terms of transparency, representability and analyzability. Trying other ranges, those lager than 25 see did rot provide enough information, and those less than 25 see id not provide additional information. Fig. 4. Daily average eee time ad bens workstation, before analysis ‘The main aspect was the bottleneck station, based on classification four clusters were defined as a patter ‘+ unfinished products, eycle time = 75 s (bad products, repair or rework necessary), ‘+ normal good products, 75s < eycle time < 175 s (products, cean be deliver on time), ‘+ normal products with short downtimes, 175 s < eyele time 300 s (good products, but not on time, minor delay),18s Péxer Dobra etal. /Procedia Manufacturing $4 2021) 184-190) ‘+ normal produets with long downtimes, 300 s < eyele time = 600 s (good produets, but not on time, major delay). the Fig. 5. Deity function and clasifcaton, bare analysis, Based on these data the bottleneck station was identified bby machine and checked by human. The human analysis reveals the sub-operations of the congestion assembly process and identify each step, determines which operation elements can be neglect or move lo anather sation, During the first line balancing (cycle 1), workstation 14 was identified as bottleneck station and the process of thread presence checking was relocated to less loaded workstations 11 (Fig. 6.) This resulted a new density function with lower eyee time of new bottleneck station (Fig. 7.) mali. Fi, 7 Density function an clnsitcation, afer iret analy In the second cyele of the hybrid analysis, the eycle time of the bottleneck was again reduced based on the same principles fiom workstation 11 to workstation 10 in two months (Fig, 8.) The key slep was to realize a not effective camera and measurement settings. Fig, 9. shows the density function after the second analysis. ig. Daily average cycle time and bateneck workstation fer second oasis. oeREEGTES Fig. 9. Density fincion and clasteto, afer second aalis When the new balanced Tine was designed then the new bottleneck station and its cycle time were determined by machine, Each process step, motion elements planned in detail by human, Manufacturing engineers have experience to realize or recognize the reasons and connections. In the case study the suitable place of process steps were decided by human, Human decision skills were combined by clustering, ‘procedure as supervised machine learning, It is important that the hybrid analysis can be applied to already operating assembly lines when there is already enough data to fulfill data mining and the work arrangement among the operators and workstations. The summary results of the two cycles are shown in Fig. 10, The bottleneck eyele time of the assembly line was reduced by 40% and the output was increased by 66%. Before applying the analysis, the OEE. value was 79% after the second cycle it increased up to 90% Without major investment Eee i ess Fig. 10, Resuts of hybrid analyse,Péter Dobra etal. /Procedia Manufacturing 54 2021) 184-190) 189 4. Conclusion ‘This article presented a hybrid analysis to identify bottleneck station at the automotive seat structure semi ‘automatic assembly Tine, The purpose of this analysis to enhance system performance and increase OEE value. Using combination of clustering as data mining method and the ‘human analysis with line balancing resulted 66% yield and 11% OBE value improvement in practice. ‘The presented analysis technique can be used with minimal investment not fonly on assembly lines, but also on individual machines ‘where a lot of data is available and the eyele time or tak time needs to be reduced for higher yields or higher OEE value, References (1) Gotan S, Hula M, Ede M, Kare H, RamsousrC, Digit Milk- Run System for a Leaming. Factory Assembly Line. Procedia Manufacturing 2019, 31:17, [2) Carlo DF, Arco MA, Tuei M. OBE Evaluation of a Paced Assembly Line tough Diferet Calculaon and Simulation Methods: A Case Sy in the Pharmuccutcal Envionment.Ttertional Joural of Engineering Business Management 214: 1-11, (3) Hedman Saramanijan M, Almstrim P. Analysis of rca factors for tomatic measurement of OFE,Procedia CRP 201687: 128-133, |) Schuh. Reinhart. Prote JP, Sauermann F, Hossthofr J, Oppolze Keol!D. Datamining dfinins ad aplication forthe managemest of ‘production comply. Proedia CIRP 201981: 374-81. [5] Bucbuer AG, Auand 58, Hughes JG. Data Minig is Manufacusag Enviromens: Goals, Techniques and Applications. Studies in Informatis and Cont 199731 (6) Kasi A. Smart manufasng.Intemationsl Jounal of Production Research 201856: SUES (7) Groger C, Niedemane F, Mivehang B. Data miaiog-diven rmanufsctriag process optinization.Proceediags ofthe Wodd Congzess oa Engincring 2012. (i) Eke J, Glss Keel A, Hanah Tisch M, Metenih J. Indusuie 40 ~ Competencies fora made producin system, rocdia Manufaure 201.25: 267-272 (9) Hoolbaler 6, BranceutherS, Reiaart G Digital Len Production ~ Ax Apeeossh to" Meni Poems for the Migration to a Digitized Production System in SMES Grom a Lean Perspective, Poveda CIRP ors: s22-827 [16] Kuve M, Hasin U, Halls M, Sidetund C, Begund M, Fin ©, Lapdsrim A. Designing Visual Managemen in Manufacturing fom a (Gar Peapective.ProcediaCIRP 209388 86-991. (UN Nakajna 8. Isteduetion to TPM: Total Pratucive Maitensnce rdutvty Pres Cambridge; 1988, 12] Olives $416, Femandes A. Costinuous provement thaigh “Lean “Tool” An Application i= a Mechanical” Company. Procalia Manathtring 201713: 1082-1088, 15] Subramanian M. Prodcton Daa Ansys ~ To idenily productivity potentials. Chalmers Univer of Technology, Gotenburg, Sweden, dos (14) Corals LC, Lamb MP, Komer MEH, Royo 2, Overall Equipment lfectivenes:Syslematic Literature Review ard Ovenaew of Dillrent Approaches. Applied Sciences 20211 6269 (15) Pasar nn 8, Hapa 1. Calealtion of OF foran Assembly Proves Toler! Journal of Reva in MechaniealEnginesing and Technolo, 20122: 25.5, (16) Antoiohh 1 Guariente P, Pein T, Raven LA, Sika FIG. Stndardvaton and epimization of an automotive _corponers production line Proved Manfacuine 201715 1120-1127, (17) Zandi S, Tabata: SAN, Ohandshary M. Evaluation of Overall [Equipment Effectiveness in a Continvous Prooa Production Sytem of CCondenste Stabilization Plant in Arslonysh, 2012; 590-598 [18 Alot 7 Pav 5, Lest V, Mona F, Jura V. Methodology of Overall auipment Elfectvenes Calculation in the Content of Industy 4.0 Enviroment Eapostasa i Nieawonote- Maintenance and Reibilty 201921: alla. [19] Hosiin i, Sarker BR. Overall uipment Eecsveness meaues of ‘gineering production system, Antal Meeting of the Decision Seiences Insitute, Conference Paper, 2046 [20] Subramaijar M, Skooph A, Salomonsson H, Bangalore P, Boast J ‘A datndeivenalgoriton to prodit throughput boleneck in production system based on active prods ofthe machines. Computers & Indus Enpnceing 2018128: 83-5 [21] Yu, Mate A, Dateeven botdenesk detection in manulacturng systems. A sainical approach, IEE otemational Conference on ‘Automaton Science and Enginectng (CASH), Tips, 2014: 710-715, [22] RonrC, Nakano M. A Quantiative Comparison of Botenerk Detection Methods ip Manufsctring Stems seth Parcular Consideration for Shiling Hotleneks. nterational Conerce on the Advances ia Production Managerent Systems, Tokyo, Japan, 2015 [23] Kamer A, Krenkther K, Schumicher Me Mathematial model for roaative rsequceing of mined todel asembly” lines” Proved Manuficturing 201933: 438245, [24] Kem W, Russchis F, Baernhas! , Planing of Workstations in Modular Automotive Atambly Syter, Procedia CIRP 2016:7:327— [25] Nguyen TL, Le MT, Va TTT, Do NIL Lean fine talansing for an ‘lectronieastemby line, Proce CIRP 201640 437-442, [26] Toksari MD, Iseyen SK, GinesE, Baykoc OF. Simple and U-type ‘Assembly Line Balancing Problems with a Learing Efe, Appa [Mathematical Modeling 200832: 2954-2961 [27] Manaye M. Line balancing techniques for productivity improvement Inverational Journal of Mechanic! and Indes) Technology 201937 95108 [28] FiManghy 1, IMaghy W. Smac Adaplable Assembly Systems Procdia CIRP 20168: 15, [29] Mora DM, Dini G. Worker Skills and Equipment Optimization in ‘Assembly Line Blmcing by a Getic Approach, Procedia CRP ove 02-107 [30] Moarzis D, Tsakalos D, Xan F, Zogopoulos V. Optimization of highly automate production line: An advanced engnering editions spproch, Proven Manufatring 2019;31 45- [SI] Pst 3, Exner ¥, Stroker N, Tanza G. Variant febity in assembly Tin Balancing under the premise of faibiy robustness, Proceia CRP 201g72: THT, [52] Sly Dtnteaing 3D product models with assembly ine balancing va process consimption Procedia Manfictrng 20187: 18-18, [33] SeirmesP. Bevezts a bg data les alkalmarisiba, Univrsis-Gy6r Nonprosit kod, Gr, 2018 [54] LaValle, Lesser, Shockey R, Hopkins MS, KruschoitaN. Big dats, ‘uals andthe path fom insights to value. MIT Sloan Managemen Review 201152: 2132 [35] Mahmood K, Lanz M, Toivonen V, Oo TA Peformance valuaton ‘Concept for Production Systems in an SME Networ. Posed CTRP 2018;72:605-608, [36] Kusik A Smart Manufocuring Must Embrace Big Duta. Nowe 2017 S84 23-25, [B97] Denkena B, Ditech MA, Wilmsmser S. Automated production data Feedback for adaptive work planing abd production convo. Proced ‘Manuficturing 2019.28: 18-28, [38] Rewer C, Brambiag F Improving daa consistency ia production ‘soul Procdia CIRP 201641 51-56, [39] Fawley W, Pits G. Knowledge Discovery i Databases. MIF Press, Camizide, United Stas, 199 [40] Harding JA, Shabbaz M, Svinivas, Kusiak A. Data mining in ‘manufichng? A review. Jounal of Muaufaeaing Science and Engacering 200,128: 9684976, [41] Chouiary AK, Harding A, Tivari MK. Datamining in manufacturing 4 review based op the kind of knowledge. Jounal of ttligest [42] Fayyad U, Piatetsky-Shepio G, Smyth P. From dat aising 10 ‘de discovery in databases, AI Magazine 199617. 37-5 [43] Canta C, AgatdB, Kusiak A peoving Manufauring Quay by Re Sequencing. Assealy Operations: A Data Mining Approach. 18k International Conference on Production Reseach aly, 2008,190 Péter Dobra et al. /Procedia Manufacturing $4 2021) 184-190 (4) Han J, Kamer M, Pei . Dat Mining ~ Concepts and Techniques, Elsevier, USA, 2012 [4] don Fiza K. Adatinyszat, Digits Tanking, 2018, [U6 Abony’ J Adsbinyiset a haldkonytigezkize, Compuerooks, Budapest, 2006 [4] Tan PN, Steinbach M, Kamar V. Inroduton to Dats Mining = Insnitr'Sotion Mal Pesteon Addison-Wesley, 2018 [4] Wirt R Hipp J. CRISP-DM: Towards» standard proces mode for dasa mining. th Procecings of the 4h Inematonal Conference on the Practical Applications of Knowledge Discovery and Data Mining 2000: 2339. [49] Huber S, Wiemee H, Schneier D, Inlet S, DMME: Data mining mehodology for enginecing applications ~ a boise extension to the ‘CRISP-DM Model, ProodiaCIRP 2019;79- 403-408, (0) Witca 1H, Ranh E.Data Mining ~ Pacical Machine Leaning Tools snd Technigues. love, USA, 2008,